1

--- Accepted by Research Policy (Dec. 2014) ---

Organizational Design of University Laboratories: Task Allocation and Lab Performance

in Japanese Bioscience Laboratories

Sotaro SHIBAYAMA a,b,*

Yasunori BABA a

John P. WALSH c,d

a University of Tokyo. Research Center for Advanced Science and Technology.

4-6-1 Komaba, Meguro-ku, Tokyo 153-8904, Japan

Tel/Fax: +81-3-5452-5371

shibayama@00.alumni.u-tokyo.ac.jp

b University of Tokyo. Department of Technology Management for Innovation.

7-3-1 Hongo, Bunky-ku, Tokyo 113-8656, Japan

c Georgia Institute of Technology. School of Public Policy.

685 Cherry Street, Atlanta, GA 30332-0345, USA.

d National Graduate Institute for Policy Studies.

7 Chome-22-1 Roppongi, Minato, Tokyo 106-0032, Japan.

* Corresponding author.

2

Abstract

A university laboratory is a fundamental unit of scientific production, but optimizing its

organizational design is a formidable task for lab heads, who play potentially conflicting roles of

manager, educator, and researcher. Drawing on cross-sectional data from a questionnaire survey

and bibliometric data on Japanese biology professors, this study investigates task allocation

inside laboratories. Results show a general pattern that lab heads play managerial roles and

members (e.g., students) are engaged in labor-intensive tasks (e.g., experiment), while revealing

a substantial variation among laboratories. Further examining how this variation is related to

lab-level scientific productivity, this study finds that productive task allocation differs by context.

In particular, results suggest that significant task overlap across status hierarchies is more

productive for basic research, and that rigidly separated task allocation is more productive in

applied research. However, optimal task allocation, with regard to scientific productivity, might

conflict with other goals of academic organizations, particularly training of future scientists. The

paper concludes with a discussion of the policy implications of these findings.

Keywords

Laboratory; task allocation; organizational design; scientific productivity; biology

3

1. Introduction

Since the modern economy relies heavily on scientific production in the academic sector,

the organizational design of academic research organizations is a critical agenda for science and

technology policy (Etzkowitz and Leydesdorff, 2000; Stephan, 1996). Academic science,

especially in natural sciences and engineering, is usually undertaken in laboratories that consist

of a lab head (also called principal investigator) and member researchers under his or her

supervision. Unlike temporary collaboration, the continuous nature of laboratories allows lab

heads with a long-range plan to set research goals, arrange a portfolio of research projects, make

investments in facilities, and accumulate and reuse a local knowledge base (Carayol and Matt,

2006; Knorr-Cetina, 1999; Latour and Woolgar, 1979; Owen-Smith, 2001). For these reasons,

prior work has suggested that a laboratory is the appropriate unit when analyzing the nature of

scientific production (Carayol and Matt, 2006; Latour and Woolgar, 1979).

Studies of the organizational design of laboratories, whether in academia or in industry,

date back to the 1950s. Among others, Pelz and Andrews (1966) examined the relationship

between scientific production and a series of organizational factors, broadly covering various

scientific fields and sectors. Subsequent literature in the sociology of science has further

investigated the roles of organizational factors such as communication, coordination, leadership,

and organizational prestige in scientific research (e.g., Allison and Long, 1990; Andrews, 1979;

Heinze et al., 2009; Hollingsworth and Hollingsworth, 2000; Long and McGinnis, 1981;

Zuckermann, 1977). Literature from other disciplinary perspectives has also advanced

understanding in specific aspects of organization; for example, the social psychology literature

studies creativity and its antecedents (e.g., Amabile, 1996) and the organization management

literature examines the motivation of researchers (e.g., Agarwal and Ohyama, 2012; Sauermann

and Stephan, 2012).

While these studies have informed how various organizational factors can affect

scientific production, they have paid limited attention to a peculiarity of university laboratories.

Academic science heavily depends on junior researchers, including students, who are often short

of experience and need training (Knorr-Cetina, 1999; Owen-Smith, 2001). Obviously,

universities are responsible not only for scientific production but also for education (Hackett,

1990), and thus, lab heads are obliged to train young members, although these two missions of

research and education could be in conflict (e.g., Fox, 1992). This is a major challenge for lab

4

heads, who have to organize the lab considering potentially incompatible goals of research and

education when deciding on task allocation for the lab head and its members. This division of

labor and potentially conflicting relationships between a lab head and members have been noted

in a few studies in the sociology of education (e.g., Delamont and Atkinson, 2001; Delamont et

al., 1997; Salonius, 2008) but analyses of their implications for scientific production have been

limited.

To fill these gaps in the literature, this study examines the organizational design of

university laboratories, highlighting the roles of lab heads and members. Investigating task

allocation in the lab context requires in-depth understanding of the distinctive activities in lab

work. In this regard, prior ethnographies of academic laboratories have illustrated in great detail

how academic science operates in one or a few specific laboratories (Knorr-Cetina, 1999; Latour

and Woolgar, 1979; Owen-Smith, 2001; Salonius, 2008). Typically, they describe task allocation

in academic laboratories as lab heads being the manager, who is busy planning, fund-raising, and

supervising members, with members being the workers, concentrating on conducting

experiments and other laborious tasks. To advance this simplified model of task allocation, we

draw on the above-outlined literature on the organization of research groups (e.g., Hollingsworth

and Hollingsworth, 2000; Pelz and Andrews, 1966; Sauermann and Stephan, 2012). In particular,

we examine two forms of possible deviation from the typical task allocation: 1) whether lab

members should engage not only in labor-intensive tasks but also in upstream tasks, and 2)

whether lab heads should engage also in labor-intensive tasks rather than staying away from the

bench like a pure manager. We argue that the optimal task allocation depends on context (Cyert

and March, 1963; Simon, 1957). In particular, we hypothesize that the pattern of task allocation

should be differentiated depending on the orientation of research in terms of being basic vs.

applied.

Drawing on interviews with 30 researchers and a questionnaire survey of 396 lab heads

from Japanese universities in the field of biology, we first draw a general picture of task

allocation in university laboratories. We find it basically consistent with the stylized view of task

allocation, but we also observe considerable variation. Second, we examine the effect of task

allocation on scientific productivity and its contingency on research orientation. Based on our

empirical results, we discuss implications for science policies.

5

2. Theory and Hypothesis

2.1. Social organization of lab work

Research activities in natural sciences are usually undertaken in laboratories that consist

of a lab head and some members under the lab head’s supervision (e.g., Carayol and Matt, 2006;

Latour and Woolgar, 1979; Owen-Smith, 2001). Lab heads are usually professors, and members

include students, postdoctoral researchers (postdocs), junior faculty, and technicians. Unlike

temporary collaboration, laboratories are characterized by a continuous form of teamwork. Lab

heads can pursue relatively long-term goals. They arrange a portfolio of research projects, some

of which may be challenging but with potentially great impact and others of which are less novel

but with limited risk, so that they can constantly produce at least minimal expected output

(Knorr-Cetina, 1999). Laboratories allow division of labor. Particularly in biology, since a

project often involves multiple techniques (Knorr-Cetina, 1999; Latour and Woolgar, 1979),

coordinating researchers with different expertise is essential. Lab tasks are also vertically divided.

Lab heads are usually responsible for setting up the research environment (e.g., funding,

equipment, and recruitment) and coordinating a series of projects, while members engage in

executing specific projects (Traweek, 1988). In addition, laboratories function as a place of

education and training. Young researchers typically consider their lab experience as an

opportunity to acquire research techniques, which will prepare them for future employment

(Delamont and Atkinson, 2001; Delamont et al., 1997).

In terms of task allocation, prior literature has mainly focused on the role of lab heads

and assumed that lab heads are occupied with upstream tasks. In a report on the career design of

American life scientists, the National Research Council (1998) mentions that “[a] principal

investigator builds a research group by defining the scientific questions to be addressed,

specifying the methods to be used, obtaining necessary funding, finding the suitable research

environment, and attracting the research personnel…. The research personnel in the group

usually work on more specific tasks that pertain to the construction of research tools or the

acquisition and analysis of data.” Similary, Knorr-Cetina (1999) finds that in the field of

molecular biology researchers often stop bench work after becoming lab heads. The role of

members, on the other hand, has been relatively understudied. A few studies in the sociology of

education, focusing on postgraduate education, have examined the division of labor between lab

heads and PhD students (Delamont and Atkinson, 2001; Delamont et al., 1997; Salonius, 2008).

6

Delamont et al. (1997), drawing on ethnographic research in British universities, suggest that lab

heads are responsible for identifying research projects and assigning them to students. Becher et

al. (1994) also point out that determining research subjects is rarely the responsibility of students.

Since mastering technical skills is the most important goal during the student’s lab experience

(Delamont and Atkinson, 2001), engaging in technical tasks seems to be regarded as the students’

primary role.

To further the discussion of task allocation, we distinguish three phases of the research

process. In general, scientific research starts from setting a research question and developing a

research plan; then, the question is tested by experiments, simulations, and other approaches; and

finally, the test results are interpreted and used to advance extant knowledge (Nightingale, 1998).

This last phase often raises new questions for future research, and the whole process is repeated.

We split this process into two phases: 1) planning, or determining research subjects and

hypotheses, and 2) execution, or testing the hypotheses, usually by experiment and data analyses

in biology. In addition, we consider the phase of 3) writing scientific papers. Planning and

execution are iterated until sufficient results are accumulated that make up a story as a paper. For

these three phases, lab ethnographies and sociology of education literature generally suggest that

lab heads are the primary player in planning and members in execution, but they are less clear

about task allocation in writing (Delamont and Atkinson, 2001; Knorr-Cetina, 1999; Latour and

Woolgar, 1979). Based on these studies, the following section first describes the general features

and rationales of task allocation for each phase. Then, we add competing arguments from the

literature on the organization of research groups (e.g., Hollingsworth and Hollingsworth, 2000;

Pelz and Andrews, 1966; Sauermann and Stephan, 2012).

2.2. Rationales of task allocation

2.2.1. Execution phase

Since biology is strongly driven by empiricism (Bertalanffy et al., 1962), biological

research heavily depends on experiments, except for purely computational or theoretical

subfields. In the execution phase, researchers attempt to transform some material substances into

interpretable information, which often takes the forms of figures or tables, and some of them are

used in the next writing phase as “results” in publications (Latour and Woolgar, 1979: Ch. 2).

This transformation may be processed manually or through devices such as DNA sequencers.

7

Some experimental procedures follow established protocols and may be made available as

commercial kits or somewhat automated, while others are less established and researchers may

have to start from developing and optimizing protocols. Importantly, even a well-established

procedure often requires substantial tacit knowledge to perform properly, such that researchers

inexperienced in that task could repeat fail to get results (Knorr-Cetina, 1999; Latour and

Woolgar, 1979). Still, lab techniques in biology are fairly standardized (Whitley, 1984), and even

PhD students can learn them in a reasonably short time. It is also noteworthy that biological

experiments usually draw on living organisms such as bacteria, cell lines, mice, and even humans

in clinical research. These organisms are highly complex and often produce unexpected results,

requiring enormous efforts of trial and error by experimenters. In addition, since living

organisms need maintenance, often on a daily basis, researchers tend to be chained to

laboratories. Experiments could take a range of times, from minutes to overnight and from weeks

to months, and researchers have to schedule their tasks depending on the life cycle of these

organisms. For these reasons, execution tasks are highly labor-intensive and time-consuming,

where members seem to have comparative advantage over lab heads (Delamont and Atkinson,

2001; Delamont et al., 1997).

Despite this stylized argument, lab heads’ co-participation could improve productivity

for a few reasons. Because lab members spend most of their time at the bench, a lab head’s

engagement in execution implies the collocation of a lab head and members. This can facilitate

communication and team coherence, both essential drivers of creativity and team coordination

(Amabile, 1996; Pelz and Andrews, 1966; Stapleton, 2004: Ch.2). Lab heads can communicate

with members through various channels such as lab meeting, but casual conversation at the

bench should lower communication barriers and may allow lab heads to obtain information

otherwise inaccessible.

More specifically, collocation enables lab heads to closely monitor members’ routine

activities. Through co-presence, lab heads could find out about and help solve problems in

members’ execution tasks in a timely fashion. This could have substantial impact on the progress

of members’ work. Biological experiment could easily take weeks or months until yielding

results, and inexperienced members could be unaware of critical mistakes that are obvious to

experienced researchers. In fact, Andrews (1979) finds that lab heads’ participation in scientific

work helps detect problems and improve team coherence. Teasley et al. (2002) show that the

8

productivity of software development projects is significantly improved when project teams are

collocated because possible bottlenecks are easily spotted and needed members are always

accessible.

Not only members’ problems but also their success are easily identified by collocation.

Biological research is often exploratory, where unintended results may bring about

breakthroughs (Merton, 1968; Merton and Barber, 2004). However, members tend to overlook

unexpected findings due to the lack of a holistic perspective and might even deliberately do so

for fear that lab heads dislike hearing unexpected news (Barber and Fox, 1958; Van Angel, 1992).

In this regard, lab heads’ close monitoring can be especially helpful. In fact, Shimizu et al.

(2012), drawing on a survey of natural scientists in the US and Japan, find that serendipitous

findings are deterred if the managerial role and the execution role are played by different

scientists.

Another rationale is technical catch-up. That is, lab heads stay able to work with the

latest lab equipment and techniques (including acquiring the necessary tacit knowledge). One lab

head we interviewed emphasized this point.

When interpreting experimental results, researchers have to distinguish true

from false signs of discoveries and to find out hidden serendipitous signs.

They may be obvious for experimenters but not for non-experimenters. It is

not rare that pure-manager lab heads misinterpret experimental results and

make silly instructions to their members. Unfortunately, members often have

to follow the instructions and tend to blindly do so especially when the lab

head is renowned.

Because most experimental techniques take a great deal of tacit knowledge, it is difficult even for

experienced lab heads to follow state-of-the-art techniques if they are not doing any experimental

work directly. Knorr-Cetina (1999) suggests that an experimental technique is a “package” of

protocol, material objects, and researchers. Lab heads may be able to understand new techniques

theoretically, but only experimenters know the knack of techniques. Lab heads will become

technically obsolete if they distance themselves from bench work (Salonius, 2008), which can be

a serious impediment when experimental techniques are rapidly advancing, as in biology.

9

2.2.2. Planning phase

Before execution, research projects must be planned. The planning phase starts with

choosing research subjects and identifying specific questions. Ultimately, the product of research

needs to be published, which is a highly competitive process (Merton, 1973). Thus, researchers

carefully choose their subjects so that they can publish more quickly and with a greater impact

than their competitors do, with prior research an essential source of information for planning.

Once research questions are settled, researchers have to translate their hypotheses into a

technically operational plan for the execution phase, which obviously requires technical

knowledge. Overall, researchers have to integrate and process various types of knowledge to

develop a feasible plan. Thus, lab heads, with greater intellectual capabilities and longer

experience, seem to have a comparative advantage for this phase (Delamont and Atkinson, 2001;

Knorr-Cetina, 1999).

The stereotyped task allocation that lab heads plan and members execute appears

reasonable considering members’ limited experience and capabilities. However, this is exactly

why they need training. Although mastering experimental skills may be the first priority for

young members (Delamont and Atkinson, 2001), learning how to design and coordinate research

projects should be indispensable. Thus, one may argue that members must be engaged in the

whole process of research for educational purposes, even if it may compromise lab performance.

In the long term, well-trained researchers should better serve scientific communities. Thus, lab

heads face the dilemma of whether to prioritize research productivity or to give their

subordinates training opportunities at the cost of productivity (Hackett, 1990). Many of our

interviewees referred to this dilemma, suggesting that there are two types of laboratories: one

where members are treated like blue-collar workers in a factory, and the other where members

are trained as future lab heads. Criticizing the former type of laboratories, one interviewee stated:

In natural sciences, it may be common that young members, especially

students, are exploited to produce experimental results as if they were

technicians. However, I believe that such an approach cannot develop good

researchers. … I believe that universities are the place for education, and thus,

students must be respected more than professors.

Although the above argument implies that members should be given training

10

opportunities in phases other than execution, it still maintains the assumption that this would

compromise productivity. However, we argue that members’ engagement in planning can

contribute to productivity. First, the creativity literature suggests that autonomy and

independence facilitates divergent thinking and exploratory problem-solving, which are critical

for creative performance (Amabile, 1996; Woodman et al., 1993). Pelz and Andrews (1966:

Ch.2) suggest that lack of autonomy is detrimental to scientific productivity, though excessive

autonomy can lead to isolation, and in particular, that sharing research goals among teammates

contributes to productivity.

Second, as often discussed in connection with autonomy, motivation is also

indispensable for creativity (Amabile, 1996; Ford, 1996; Pelz and Andrews, 1966: Ch.6). Roach

and Sauermann (2010) find that PhD students in science and engineering show strong preference

for freedom to choose research projects. Even inexperienced members desire to be involved in

decision making. If autonomy in project selection is given, members can ascribe their success or

failure to their own actions, which stimulates their intrinsic motivation (Hackman and Oldham,

1976). Thus, involving members from the outset of the research process can encourage them to

seriously engage in later phases, but isolation from the planning phase can demotivate them. One

lab head we interviewed argued:

Because the execution phase in life science research is painstakingly

laborious, members could not go through it without strong intrinsic

motivation. In this regard, having members engage in planning is effective. I

try to respect members’ choice of research topics even if they seem likely to

fail, hoping for them to reach a serendipitous discovery.

Third, members’ engagement can contribute to coordination across phases, which,

coupled with autonomy, is critical for scientific productivity (Pelz and Andrews, 1966: Ch.2).

Though the three phases may appear to proceed linearly, upstream tasks are often revised based

on feedback from downstream tasks. In the planning phase, research plans may have to be

adjusted in accordance with experimental results. If members, who are the main player in the

execution phase, are involved in the planning phase, they should be able to understand the aim of

execution tasks, efficiently inform lab heads of experimental results, and properly fulfill changed

plans.

11

2.2.3. Writing phase

Writing is the process of producing a scientific paper for publication using the results

obtained in the execution phase. Although the process of writing a paper may span all three

phases in that literature may be reviewed during planning and figures and tables are created

during execution (Latour and Woolgar, 1979: Ch. 2), we narrowly define it as the process of

writing after results are produced. Still, writing is more than a mechanical process of

summarizing experimental results and can be an intellectual process of interpreting results,

placing them adequately in the context of concurrent scientific debates, and creating a story that

interests peers. These tasks take a set of skills. First, since biological research projects usually

involve multiple members with different technical expertise, the execution phase produces

various types of results, among which appropriate ones need to be selected, as the first step of

writing. This coordination role seems manageable only by lab heads who have the authority and

a holistic viewpoint beyond each member’s. Second, biological research is often serendipitous,

and experimental results are unpredictable (Merton, 1968; Merton and Barber, 2004; Whitley,

1984: Ch.4). To write a paper with unexpected results, authors may have to start over from

literature review and revise the original storyline. Thus, the quality of papers can be greatly

affected by authors’ theoretical knowledge. In addition, the ability to find out serendipitous

results depends on the observer’s knowledge (Seymore, 2009). As Pasteur said, “Fortune favors

the prepared mind.” Lab heads may be better able to see the potential in experimental results.

Third, the writing process can involve informal communication with other researchers. Before

submitting a paper for peer review, authors attempt to increase the likelihood of acceptance by

incorporating the knowledge of leading researchers in the field. Some of our interviewees

emphasized that negotiations with journal editors are also indispensable. Thus, this phase seems

to require both intellectual and social skills that are better exercised by experienced lab heads

than by young members.

However, this argument depends on how much value needs to be added in this phase.

For example, if the research goal is practical, not much theoretical knowledge may be

additionally required. If experimental results are predictable, the storyline of a paper can be fixed

before execution and the value-added in this phase can be limited (Whitley, 1984). Then,

members’ writing may be justified since their labor cost is lower. In addition, assuming that

12

members are the main player of execution, lab heads’ writing can be inefficient because it incurs

communication costs and might suffer from imprecise or incomplete knowledge transfer. Thus,

members may be able to write papers more efficiently than lab heads under some circumstances.

2.3. Contextual contingency

The previous section outlines competing rationales for task allocation in each phase. The

overall impact of task allocation on scientific productivity is thus determined by the balance of

these rationales, which we argue is contingent on organizational context. Prior literature has

identified various contextual factors that explain scientific productivity, such as leadership,

organizational prestige, size, and age (Allison and Long, 1990; Heinze et al., 2009; Long and

McGinnis, 1981; Pelz and Andrews, 1966). We suppose that these factors not only directly affect

scientific productivity but also can change the effect of task allocation on scientific productivity.

Among other contextual factors, we formulate hypotheses on contingency on research

orientations, comparing basic vs. applied research.

Biology is a broad discipline and is related to several research fields such as medicine,

agriculture, and pharmaceuticals. This diversity in research goals can be relevant in coordinating

lab work (Sauermann and Stephan, 2012; Whitley, 1984). Some researchers seek general

understanding of certain phenomena while others are guided by consideration for practical use,

where the former is called basic and the latter applied (Stokes, 1997). Calvert (2004) contrasts

basic and applied research, identifying their characteristics. First, basic research is unpredictable,

where researchers aim to find a new concept or push the boundaries of existing knowledge. This

feature in basic research leads to an exploratory approach compared to a more confirmatory

approach in applied research. Second, basic research is general in that results can be used for a

wide range of instances and phenomena while applied research helps solve a specific problem.

Third, basic research is driven by the theoretical dynamics inside the discipline. This is also

related to the generality as theories involve statements of general principles. We assume that

these features of basic or applied research affect the productivity of different task allocations.1 In

what follows, we discuss how research orientations change the impact of different task

1 Research areas may be a result of strategic choice, but we assume that this choice is less dynamic or occurs only in

the longer term. For example, the basicness of research is substantially affected by disciplines, and thus, by the

department laboratories belong to. Thus, in the short term, research areas could be regarded as a given contextual

factor.

13

allocations in each phase.

2.3.1. Member’s engagement in planning

In the planning phase, we have referred to autonomy, motivation, and coordination as

the rationales of members’ engagement. We argue that these factors play a more important role in

basic research than in applied research. Intrinsic motivation and autonomy are regarded as

antecedents of creativity in general. The mechanism behind this argument is that they facilitate

exploratory and divergent thinking (Amabile, 1996), which is more relevant to basic research

(Calvert, 2004). For example, Sauermann and Cohen (2010), based on a survey of industry

researchers, show that intrinsic motivation contributes to productivity to a greater extent in

upstream R&D activities than in downstream. In addition, the exploratory nature of basic

research implies that plans in basic research are prone to frequent updating. Based on

experimental results, researchers have to frequently adjust their research plan; thus the feedback

loop between planning and execution should be tightly linked (Nelson, 1959). This can be

streamlined by members’ engagement in the planning phase. In contrast, this potential benefit

seems limited in applied research, where the goal of research is clearer and members can stick to

the original plan. Thus, we hypothesize:

Hypothesis 1: Members’ engagement in planning has a more positive effect on productivity

in basic research than in applied research.

2.3.2. Lab head’s engagement in execution

We have pointed out collocation and technical catch-up as possible justifications for lab

heads’ engagement in execution, which we argue are particularly important in basic research. As

discussed above, the exploratory nature and abstract goals in basic research imply that research

plans tend to be loosely predetermined and frequently updated. Exploratory research encourages

autonomous trials and errors, but this incurs the risk that members are stuck in trivial problems

or unpromising lines of research. This calls for good communication between a lab head and

members and lab heads’ direct monitoring even though it is costly. Furthermore, since the

success of basic research depends more on unplanned findings, a keen eye for serendipitous

signs in experimental results is essential. Given that young members have limited capabilities in

this respect, lab heads’ collocation may be needed. Finally, to effectively monitor members and

14

detect interesting signs from noisy experimental results, lab heads need to maintain their

technical expertise by catching up with the latest technologies. In sum, we hypothesize:

Hypothesis 2: Lab heads’ engagement in execution has a more positive effect on

productivity in basic research than in applied research.

2.3.3. Task allocation in writing

We have discussed that the optimal task allocation in writing depends on the balance

between the expected value-added in writing and labor cost, which we argue differ by research

orientations. The ultimate goal of applied research is, by definition, application (Calvert, 2004;

Stokes, 1997). Thus, practically useful findings can be valued and published even if they do not

advance theoretical understanding. For example, research on clinical medicine can be published

if it proves the efficacy of a drug substance but does not elucidate its mechanism. On the other

hand, basic research aims to advance knowledge and tends to refer to general and abstract

concepts (Calvert, 2004; Stokes, 1997). Thus, researchers have to understand the up-to-date

theoretical debate and incorporate their findings in it. In this regard, lab heads’ advantage over

young members in writing is greater in basic research.

The unpredictable nature of basic research could foster this tendency. Basic research

takes a more exploratory approach and applied research a more confirmatory approach (Calvert,

2004). Basic research often starts from a broad question without having a precisely testable

hypothesis, and experimental results might be applied to a diverse range of scientific discussion.

If researchers have a broad range of knowledge not only about the originally intended areas but

also about surrounding areas, serendipitous discoveries are more likely. In this regard, writing in

basic research can be more a creative process of generating a novel story. For example, this

interviewee suggested the necessity of substantial knowledge and experience in writing in basic

research.

Serendipitous discoveries are important in biological, particularly basic,

research. I think that even young researchers could find unintended results if

they are careful enough. However, it does not guarantee publication. For

publication, serendipitous discoveries must be theorized and proved in

accordance with the extant theories. This takes substantial knowledge and is

not feasible for inexperienced researchers.

15

On the other hand, applied research tends to have a clear focus on application (Calvert, 2004). To

the extent that the research goal is specific, room for creative interpretation is limited. Thus,

writing in applied research can be a process of summarizing experimental results according to

the predetermined plan. In this case, inexpensive members’ labor lends them a comparative

advantage. In summary, we hypothesize:

Hypothesis 3: Lab heads’ engagement in writing has a more positive effect on productivity

in basic research than in applied research, while members’ engagement in writing has a more

positive effect on productivity in applied research than in basic research.

3. Data & Method

3.1. Sample and data

To test our hypotheses, we conducted a questionnaire survey of lab heads of Japanese

biology laboratories and collected their publication data from the Web of Science (WoS), which

yielded cross-sectional data of 396 laboratories.

To begin, we interviewed 30 Japanese researchers to investigate the context of

university laboratories and to design the survey instrument. The sampling frame of the survey

was prepared with the following criteria. First, we chose researchers currently in the position of a

full professor. Japanese universities have a three-level promotion system with full professors at

the top, followed by associate professors, and then, by assistants or lecturers. Before becoming

an assistant or lecturer, researchers tend to experience a few years as a postdoc. Typical biology

laboratories consist of senior staff (full and/or associate professor), who are the lab head, and

members, including a few junior faculty members (assistant or lecturer), postdocs, students, and

technicians. Unlike American universities, junior faculty members are often under the

supervision of lab heads. Some associate professors have independent laboratories, and others

work in the same laboratory with a full professor, often co-supervising a laboratory. Second, we

chose researchers who have received a national research grant in the field of biology at least once

in the previous three years (2007-2009), which implies that they are active researchers.2

2 We prepared our sampling list using the government’s database of Grants-in-Aid for Scientific Research (GiA)

(https://kaken.nii.ac.jp/en/). This sampling strategy is based on the assumption that academics who received no grant

for three years are not researchers. Although doing research without receiving this particular grant is possible,

previous research shows that it is not common (Shibayama, 2011), as GiA is the primary funding source for

16

Drawing on the list of grant recipients, we prepared a sampling frame of 1,378 lab heads. After

re-examining their research fields and affiliations with public information, we identified 900 lab

heads in 56 universities as a final sample. We mailed the survey and collected 396 responses

(response rate = 44%) after two waves of requests, May through July 2010.3

We collected bibliometric data for the 396 respondents. We primarily drew on

publications in 2007-2011 to measure the current productivity of each laboratory. Because lab

heads in biology usually become an author of all the papers produced in their laboratories,

publications including lab heads as an author should cover most publications from their

laboratory. As search criteria for the WoS, we used the names and affiliations of the respondents

and research areas being in life sciences. After downloading all matched data, we excluded false

matches on the basis of authors’ full names, affiliations, and so forth.4

3.2. Measurement

Lab productivity. We prepared two measures of productivity at the lab level. First, we

counted the number of publications authored by our respondents in 2007-2011 (pub count).

Second, to measure a more quality-focused aspect of scientific production, we drew on citation

counts. To address the age effect of citation, we summed up citation counts using a three-year

window for each of the papers (citation count).5 Then, we divided pub count and citation count

by the number of lab staff (lab heads and junior researchers) and use their logarithms in

regressions.6 Because these two measures are highly correlated and the regression results are

similar, we mainly report the results based on the citation count.

Organizational structure and task allocation. We asked respondents about the number

of senior staff (full and associate professors), junior researchers (assistant professor, lecturer, and

Japanese academics and widely awarded (http://www.jsps.go.jp/j-grantsinaid/index.html). 3 To examine non-response bias, we randomly selected 50 non-respondents and found no significant difference

between the response and non-response groups in publication productivity, organizational rank, or gender (p > 0.1). 4 We assumed that authors whose full name and affiliation are the same are identical. Since the majority of

publication data in 2007-2011 include author full names, we first excluded false hits whose authors share only

initials but not full names. Approximately 20% of our publication data in 2007-2011 do not have author full names,

and we determined whether their authors were our respondents or not by comparing cited references and coauthors’

names with publication data whose author full names are available. 5 Since newer publications tend to have fewer citations, we fix the time window of citation at three years (i.e.,

citations within three years after publication are counted). We drew on publications up to 2011 for the same reason:

citation counts would be unreliable for too recent publications. 6 We also computed lab productivity using fractional count of publications and citations based on the number of

authors (Lindsey, 1980). The results are qualitatively similar (Table S1 in Supplementary Results), so we present

results with ordinary measures.

17

postdocs), PhD students, and technicians in each laboratory. The total number of these

researchers is used as lab size. On average, a laboratory consists of 1.6 senior staff, 2.6 junior

researchers, 2.8 PhD students, and 1.1 technicians. To examine task allocation, we define six

research tasks: 1) choosing a subject, 2) formulating a hypothesis, 3) planning experiment, 4)

doing experiment, 5) analyzing data, and 6) writing papers. We argue that 1) – 3) corresponds to

the planning phase, 4) and 5) to the execution phase, and 6) to the writing phase. For each of

these six tasks, we asked the extent of involvement of full professors (our respondents), associate

professors, junior researchers, and PhD students, respectively. The responses take a three-point

scale: 0: no role, 1: minor role, and 2: main role, where we allow a “main role” to be played by

more than one rank group. When a laboratory is co-supervised by full and associate professors,

we take the maximum value of their engagement as the lab head’s engagement. For regression

analyses, we prepared the following three variables corresponding to the three hypotheses

respectively. First, we averaged the six items regarding the planning of junior researchers and

PhDs (members’ planning). Second, we averaged the two items on lab head’s engagement in

execution (lab head’s execution). Third, we took the average of researcher’s writing, PhD’s

writing, and the inverse of lab head’s writing (members’ writing (& lab head’s not writing)). In

the following analyses, we focus on laboratories that have at least one junior researchers and one

PhD student to consistently compute the task allocation measures.7 Of the 396 laboratories, 305

laboratories (77%) satisfy this condition with a mean size of 9.2 and the standard deviation of

5.2.

Basic research. To measure the research orientation of each laboratory, we asked

“which describes your research goal, basic (aiming at advancement of theory and knowledge) or

applied (aiming at solving problems in the real society)?” with a five-point scale, 1) mostly basic,

2) more basic than applied, 3) both to a similar extent, 4) more applied than basic, and 5) mostly

applied. Of our respondents, 55% chose 1), implying that their research goal was completely

basic. For these laboratories, a dummy variable is coded one, and other laboratories are regarded

7 For our measurement design, we cannot properly compute the task allocation variables in laboratories without

junior researchers or PhDs. Since excluding these laboratories could cause a bias, we conducted a sensitivity

analysis to confirm that it does not affect our results (Table S2 in Supplementary Results). We further analyze the

nature of the dropped laboratories and find that they are significantly smaller than laboratories included. By

additionally examining the recent career of all lab heads, we also find that the lab heads of dropped laboratories are

more likely to retire soon after our survey. Lab heads about to retire tend to avoid employing members to be ready

for closing their laboratories.

18

as applied with the dummy coded zero (basic research). To validate this subjective measurement,

we drew on the type of journals where the respondents publish their papers. Based on a

classification of the “basicness” of journals (Narin et al., 1976), we calculated the percentage of

respondents’ papers in basic journals and confirmed that basic laboratories tend to publish in

basic journals (r = .41, p < .001). Second, we surveyed the number of patent applications in

2009–2010 and confirmed that basic laboratories have significantly fewer patents than applied

laboratories (.31 vs. .85 applications per year; p < .001). Third, we identified the research field in

which each respondent received the majority of his national grants. We categorized these fields

into basic and applied fields8 and confirmed that this measure is correlated with basic research (r

= .41, p < .001). In addition, we tested the assumption that basic research is more exploratory

and applied research is more confirmatory (Calvert, 2004). We surveyed “which describes the

quality of your research, exploratory or confirmatory?” with a similar five-point scale, finding

that this is positively correlated with the measurement of being basic vs. applied (r =.23, p

< .001).

Control variables. The productivity of individual researchers should affect both lab

productivity and task allocation. To incorporate lab head’s individual productivity, we drew on

publications authored by the respondents during the five years before they obtained a tenured

position (i.e., before they opened their own laboratory). For these publications, similarly to the

lab productivity measures, we computed publication count and citation count with a three-year

citation window (pre-tenure pub count and pre-tenure citation count). As a measure of

organizational prestige, we prepared a dummy variable for the top seven pre-imperial universities

(top 7 univ) because they enjoy exceptionally high prestige both in research and in education

among other Japanese universities (Kneller, 2007).9 As measures of research input, we include

per-staff research budget (JPY in million: budget/#staff) and lab head’s average hours spent on

research activities (time for research). Time for research was measured with a six-point scale, 1)

less than 10 hours, 2) 10-20 hours, 3) 20-30 hours, 4) 30-40 hours, 5) 40-50 hours, and 6) 50

hours or longer per week. Several measures for individual attributes are included. We controlled

8 Basic fields include basic medicine, neuroscience, genome science, etc. while applied fields include agricultural

science, pharmaceutical, clinical medicine, etc. 9 As of 2010, Japan had 778 national, regional and private universities with four-year degree programs

(http://www.e-stat.go.jp/). Among them, national universities are the primary player in academic research while

most private universities are education-oriented. Among 86 national universities, the top seven (Universities of

Tokyo, Kyoto, Osaka, Tohoku, Hokkaido, Kyushu, and Nagoya) are designated as pre-imperial (Kneller, 2007).

19

for the number of years since lab heads opened their laboratory (lab age). We asked about

experience of research abroad with a six-point scale, 1) none, 2) less than half a year, 3) one year,

4) 2 years, 5) 3 years, and 6) 4 years or more (foreign experience). If the current laboratory is

where they obtained their degree, a dummy variable is coded one (inbred). If a lab head had a

medical doctor degree, a dummy variable is coded one (medical doctor). If a lab head is female,

a dummy variable is coded one (female). In addition, we identified the research field of each

laboratory on the basis of fields attributed to the lab heads’ past research grants, and in regression

analyses, we included 21 dummy variables for these fields.10

Table 1 presents the descriptive

statistics and correlation matrix of these variables.

<< Insert Table 1 about here >>

4. Results

4.1. Variation in task allocation

Figure 1 illustrates the extent of engagement in six research tasks by three ranks. It

indicates that the planning phase is primarily conducted by lab heads and execution by members

(junior researchers and PhDs). This confirms the stylized model of task allocation (National

Research Council, 1998; Delamont and Atkinson, 2001; Knorr-Cetina, 1999). However, the

extent of lab heads’ engagement in execution and that of members’ planning show substantial

variation, suggesting that the division of labor is not uniform across labs. As for the writing

phase, lab heads are highly committed but members are also engaged; only about 10% of

members play no role in writing.

<< Insert Figure 1 about here >>

Then, we analyze the extent of deviation from the typical task allocation in Table 2. In

each phase, we identified laboratories where the main role of each phase is played by the atypical

rank. For simplicity, we took the average of junior researchers’ and PhDs’ engagement as

members’ engagement. In about one-third of the laboratories, members are engaged in planning

(Row 1), and in about a half of the laboratories, lab heads are engaged in execution (Row 2) and

members are engaged in writing (Row 3). Again, these results suggest that the division of labor is

10

Fields include neuroscience, environmental science, agricultural science, pharmaceutical, basic medicine, clinical

medicine, structural biochemistry, biophysics, molecular biology, cell biology, and so forth.

20

not uniform. In particular, we do not see a strict division of labor such that lab heads plan and

members execute. Further, we break down the patterns of task allocation into the eight

possibilities of three rank-phase combinations (Rows 4-11). Rows 4 and 5 are supposedly a

typical pattern, which account for 35%, but other patterns are not rare (12-20%) except Rows 8

and 10.

<< Insert Table 2 about here >>

Finally, to validate our three variables of task allocation, we run a factor analysis and a

varimax rotation for the 18 measures of task allocation (three ranks x six task types). We obtain a

six-factor solution based on the Kaiser-Guttman criterion (i.e., eigenvalues greater than one).

Appendix 1 presents the factor loadings of the six factors. This solution is consistent with our

instrument design in that tasks intended to be in the same phase are actually attributed to the

same factor. For example, Factors 1 and 3 highlight the measures for three planning tasks, and

Factors 4 and 5 for two execution tasks (indicated in bold italic). Furthermore, our three

variables of task allocation are found in three distinctive factors; Factor 1 corresponds to

members’ planning, Factor 5 to lab head’s execution, and Factor 6 to members’ writing (and lab

head’s not writing).

4.2. Prediction of scientific productivity

Tables 3 and 4 present the results of regression analyses with per-staff citation count and

pub count as the dependent variables, respectively. For citation count (Table 3), Model 1, based

on the whole sample, shows a weakly positive effect for members’ planning (b = .241, p < .1)

and insignificant effects for lab head’s execution and members’ writing (p > .1). As for control

variables, pre-tenure citation count shows strongly positive effects in all models and female

shows strongly negative effects in Models 1-3. To investigate the contextual contingency, we

split the sample into basic and applied laboratories (Models 3 and 4). Members’ engagement in

planning shows a significantly positive effect in basic laboratories (b = .476, p < .05) and a

positive, though insignificant, effect in applied laboratories (b = .280, p > .1). Interestingly, lab

head’s execution shows opposite signs between two subsamples: positive in basic laboratories (b

= .481, p < .05) but negative in applied laboratories (b = -.385, p < .05). This implies that the cost

of lab head’s execution might exceed its benefit in applied laboratories. Finally, members’

21

writing shows a significantly negative effect only in basic laboratories (b = -.527, p < .1). In

order to statistically compare the coefficients between basic and applied laboratories, we draw on

two approaches. First, we add interaction terms for task allocation and basic research in Model 2

for the whole sample. The results show a strongly significant interaction effect for lab head’s

execution (b = .895, p < .001). The other two interaction terms are insignificant although their

signs agree with our expectation. Second, we directly compare the coefficients between Models 3

and 4, allowing the coefficients of all independent variables to differ between the two

subsamples.11

The result does not show statistical difference for members’ planning (p > .1) but

show strongly significant difference for lab head’s execution (p < .001) and weakly for members’

writing (p < .1). These results support Hypothesis 2 and weakly supports Hypothesis 3, but do

not support Hypothesis 1.

Table 4 shows qualitatively similar results with publication count as the dependent

variable. Comparing basic and applied laboratories, Model 2 shows significant interaction effects

for lab head’s execution (b = .429, p < .001) and for members’ writing (b = -.327, p < .1). In

addition, direct comparison between Models 3 and 4 show significant differences for all three

task allocation measures (p < .1, p < .01 and p < .05, respectively). Thus, all the hypotheses are

supported when publication count is used as the dependent variable.

The contrast between Tables 3 and 4 suggests that the effect of task allocation could be

different for the qualitative and quantitative aspects of scientific production. Figure 2 illustrates

the association between task allocation and the two productivity measures in terms of the three

task allocation measures. It suggests that in applied laboratories members’ planning may

contribute to the quality (citation count) but not to the quantity (pub count) of publication, and

likewise, members’ writing may increase quantity but may not matter for quality.

<< Insert Tables 3 and 4 and Figure 2 about here >>

5. Discussion and Conclusions

5.1. Findings and implications

Drawing on a sample of Japanese biology laboratories, this study first examines the

patterns of task allocation. While our results confirm the prior assumption that lab heads

11

Namely, we included interaction terms with basic research for all independent variables (not only focal variables

but also other control variables). In addition, we drew on seemingly unrelated estimation technique with the STATA

command suest (Weesie, 1999). Both methods yield qualitatively similar results.

22

primarily engage in planning tasks and members in execution tasks (Delamont and Atkinson,

2001; Delamont et al., 1997; Salonius, 2008), they also show some variation in task allocation.

These results contribute to drawing a general picture of organizational design of academic

laboratories, developing prior literature based on ethnographies (Knorr-Cetina, 1999; Latour and

Woolgar, 1979; Owen-Smith, 2001; Salonius, 2008).

Second, this study investigates how variation in task allocation affects scientific

productivity in different contexts. To this end, we build a framework of three research phases

drawing on lab ethnographies (e.g., Knorr-Cetina, 1999; Latour and Woolgar, 1979; Salonius,

2008). Furthermore, we discuss the rationales of task allocation between lab heads and members

based on literature on the organization of research groups (e.g., Hollingsworth and Hollingsworth,

2000; Pelz and Andrews, 1966; Sauermann and Stephan, 2012). Our data highlight productive

patterns of task allocation in each research phase. In planning, although lab heads are the primary

decision makers (Delamont and Atkinson, 2001; Knorr-Cetina, 1999), members’ participation

does contribute to productivity, possibly because autonomy stimulates members’ intrinsic

motivation (Amabile, 1996; Roach and Sauermann, 2010) and may encourage their effort in later

phases. Comparing basic and applied research, members’ engagement in planning is more

important in basic research due to its exploratory nature and the relevance of intrinsic motivation.

In execution, members are believed to be the primary player because lab heads are too expensive

for labor-intensive tasks (Delamont and Atkinson, 2001; Delamont et al., 1997). However, we

hypothesize that the cost of lab head’s labor can be justified by the benefit of collocation and

technical catching up (Andrews, 1979; Teasley et al., 2002). Our results confirm this hypothesis

in basic laboratories, possibly because sharing workspace with members, having frequent

discussion, and updating research plans in a timely fashion are essential for exploratory research.

In contrast, the cost seems greater than the benefit in applied research, where the research is

more confirmatory and more likely to follow predetermined plans. In writing, the results show

that lab heads but not members should be the primary player, particularly in basic laboratories.

Since basic research is more theory-driven and exploratory (Calvert, 2004), lab heads may better

serve writing tasks with their longer experience and holistic scientific perspective.

Academic laboratories are peculiar in that they are responsible not only for research but

also for education. However, these two goals can be incompatible (e.g., Fox, 1992; Hackett,

1990) and cause a dilemma for lab heads, who may have to prioritize either members’ training or

23

scientific productivity at the cost of the other. Indeed, our results suggest that task allocation

optimized for research productivity can be different from that for education. For example, even

though young members should be taught how to write a paper, our results suggest that it

compromises productivity in basic research. This conflict between research and education may

be becoming more serious as science policies increasingly emphasize research productivity and

researchers have been under pressure for short-term evaluation based on publications. Obviously,

the training of future researchers is indispensable in sustaining the science system. Therefore,

science policies should take actions to better balance research and education.

Interestingly, our results also suggest that research productivity and training members

can be partially compatible by showing that members’ participation in planning improves

productivity. Nevertheless, as the stylized task allocation assumes (Delamont and Atkinson,

2001; Knorr-Cetina, 1999; Latour and Woolgar, 1979), we observe that members are not engaged

in planning in many laboratories. This may be because of the hierarchical lab structure in Japan,

modelled on the German chair system (Arimoto, 2011), where a vertical division of labor is clear,

in comparison to flatter organizations observed in the US and other Western countries (Kneller,

2007). If this is the case, science policies might need to restructure the lab design or to change

employment practices to allow greater participation for junior members. In fact, the Japanese

government has embraced such a vision (Kneller, 2007), but it seems yet to be realized.

5.2. Future directions and limitations

This section discusses some potential directions of future research and limitations of the

current study. First, this study focuses on research orientations as a contextual factor, but other

organizational contexts can affect the optimal form of organizational design. For example, we

also explored the impact of organizational prestige, lab size, and lab age (Heinze et al., 2009;

Levin and Stephan, 1989; Tunzelmann et al., 2003), finding some contingency effects.12

For

example, we find that members’ planning has a stronger effect in prestigious universities than in

non-prestigious universities, probably because the ability of students is correlated with university

prestige due to the nature of the admission system in Japan (Kneller, 2007). We also find that lab

head’s writing is more effective in small laboratories and in old laboratories, especially in basic

research. The former may be because lab heads in large laboratories cannot efficiently handle the

12 Table S3 in Supplementary Results

24

writing task for too many members (Delamont and Atkinson, 2001; Salonius, 2008). The latter

may suggest that accumulated knowledge in senior lab heads lends them advantage in writing.

These results imply that the contingency of task allocation should be investigated with multiple

contextual factors taken into account, for which more robust tests are needed in future research.

Second, although we analyze the contingency of task allocation by examining the

interaction effects of research orientation and task allocation, a different form of contingency is

plausible. That is, if lab heads know the optimal forms of task allocation under a certain context,

they should be able to adapt their organizational design to the context (Drazin and Vandeven,

1985). Tables 1 and 2 imply that this might be the case. For example, we find that members’

writing is more common in applied laboratories than in basic laboratories. Thus, lab heads in

applied laboratories may understand the benefit of members’ writing and actually have them

write. On the other hand, the results also suggest that the productive task allocation is not always

followed. For example, although lab head’s execution increases productivity in basic laboratories

and decreases it in applied laboratories, actual task allocation shows no sign of adaptation. Since

these results are limited due to the nature of our cross-sectional data, future research should draw

on dynamic analyses to further the understanding of adaptation.

Third, this study focuses on the laboratory as a unit of analysis, but a laboratory usually

engages in multiple projects, and project can be regarded as another important unit of analysis in

examining scientific production. It is plausible that task allocation in one project is different from

that in another. We attempt to address some potentially confounding factors attributed to projects

such as project size, external collaboration, and interdisciplinarity.13

Controlling for these factors,

we confirm that our results remain qualitatively unchanged. Nevertheless, future research is

needed to closely inspect the attributes of projects in addition to those of laboratories.

Fourth, since this study draws on a specific sample of Japanese biology laboratories,

further research is needed for generalization. Countries can differ in the degree of emphasis on

the practical application of academic science and in the practice of PhD training, both of which

are related to the focal concepts of this study. For the former, a few studies have shown

comparable figures between Japan and the US regarding academics’ practical orientation (e.g.,

patenting, commercial activities) (e.g., Shibayama et al., 2012). For the latter, statistics of the

OECD (2013) suggest, for example, that comparable proportions of PhD graduates are employed

13

Table S4 in Supplementary Results

25

in industry in the US, Japan, and some European countries. Although these lend some confidence

to the representativeness of our sample, it is still possible that our results are specific to the

context. The generalizability of our results from biology to other fields of science also needs

further investigation.

Fifth, we cannot fully address the issue of endogeneity. It is plausible that task allocation

is determined by the ability of lab heads and members. In this regard, members’ ability is

difficult to measure. From our results, we assume that the pattern of task allocation is determined

by lab head’s policy and preferences14

and is rather stable over time and across projects, but

longitudinal analyses are needed for more robust tests of causality in future research.

Sixth, we made a dichotomous distinction between basic and applied research, but

research area is more continuous in reality. A laboratory can engage in both research areas. In

particular, since recent science policies emphasize practical application (Etzkowitz and

Leydesdorff, 2000), basic laboratories are under pressure to engage also in more applied research.

Stokes (1997) suggested that a basic-applied combinatorial approach (so-called Pasteur’s

Quadrant) is fundamentally different from pure basic and pure applied. In this regard,

project-level analyses may be helpful. Related to this point, our argument is based on the

assumption that research areas are a predetermined context rather than a result of strategic choice.

We believe that this is an acceptable assumption in the short term. However, future research

should investigate the dynamics between organizational structure and the longer-term trajectory

of research areas.

5.3. Concluding remark

This study presents evidence on the division of labor in Japanese biology laboratories,

and shows that while the stereotype of the lab head planning and the members executing is not

rare, other combinations are also quite common. In addition, these “off-diagonal” labs can be

more or less effective than the “normal” structure, depending on the research orientation. These

findings suggest a variety of follow-up questions about optimal task allocation, the need to

balance the sometimes conflicting goals of research productivity and effective education, and the

tensions between intrinsic motivations and efficient production. Our findings also suggest the

14

To investigate lab heads’ training policy and how it affects the pattern of task allocation, we asked several

questions on lab head’s policies regarding student training, and we observed significant correlations between lab

head’s policies and task allocation patterns (Table S5 in Supplementary Result).

26

utility of taking an organizational studies approach to the analysis of scientific labs. Such an

approach can help generate new research questions and contribute important results to the

science of science policy.

Acknowledgments

We acknowledge financial support from the Konosuke Matsushita Memorial Foundation,

Grant-in-Aid for Research Activity Start-up of the Japan Society for the Promotion of Science

(#23810004), and Grant-in-Aid for Scientific Research (B) Program (#20330077) form The

Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan. We

acknowledge the U.S. National Science Foundation and the Patent Board for offering us the

journal classification data. We appreciate thorough and encouraging comments from three

anonymous referees. We acknowledge Ms. Aya Sasaki for technical support.

References

Agarwal, R., Ohyama, A. 2012. Industry or academia, basic or applied? Career choices and

earnings trajectories of scientists. Management Science.

Allison, P. D., Long, J. S. 1990. Departmental effects on scientific productivity. American

Sociological Review, 55: 469-478.

Amabile, T. M. 1996. Creativity in context: Update to the social psychology of creativity.

Boulder, CO: Westview Press.

Andrews, R. 1979. Scientific productivity: The effectiveness of research groups in six

countries. Cambridge: Cambridge University Press.

Arimoto, A. 2011. Japan: Effects of changing governance and management on the academic

profession. In Locke, W. & Cummings, W. & Fisher, D. (Eds.), Changing governance

and management in higher education: The perspectives of the academy: 281-320.

Dordrecht: Springer.

Barber, B., Fox, R. C. 1958. The case of the floppy-eared rabbits: An instance of serendipity

27

gained and serendipity lost. American Journal of Sociology, 64: 128-136.

Becher, T., Henkel, M., Kogan, M. 1994. Graduate education in britain. London: Jessica

Kingsley.

Bertalanffy, L. V., Rapoport, A., Meier, R. L. 1962. General systems yearbook. Washington,

D.C.: Society for General Systems Research.

Calvert, J. 2004. The idea of 'basic research' in language and practice. Minerva, 42: 251-268.

Carayol, N., Matt, M. 2006. Individual and collective determinants of academic scientists'

productivity. Information Economics and Policy, 18: 55-72.

Council, N. R. 1998. Trends in the early careers of life scientists. Washington, DC: National

Academy Press.

Cyert, R. M., March, J. G. 1963. A behavioral theory of the firm. Englewood Cliffs, NJ:

Prentice-Hall.

Delamont, S., Parry, O., Atkinson, P. 1997. Critical mass and pedagogic continuity: Studies in

academic habitus. British Journal of Sociology of Education, 18: 533-549.

Delamont, S., Atkinson, P. 2001. Doctoring uncertainty: Mastering craft knowledge. Social

Studies of Science, 31: 87-107.

Drazin, R., Vandeven, A. H. 1985. Alternative forms of fit in contingency theory. Administrative

Science Quarterly, 30: 514-539.

Etzkowitz, H., Leydesdorff, L. 2000. The dynamics of innovation: From national systems and

"mode 2" to a triple helix of university-industry-government relations. Research

Policy, 29: 109-123.

Ford, C. M. 1996. A theory of individual creative action in multiple social domains. Academy of

Management Review, 21: 1112-1142.

Fox, M. F. 1992. Research, teaching, and publication productivity - mutuality versus competition

in academia. Sociology of Education, 65: 293-305.

Hackett, E. J. 1990. Science as a vocation in the 1990s - the changing organizational culture of

academic science. Journal of Higher Education, 61: 241-279.

28

Hackman, J. R., Oldham, G. R. 1976. Motivation through the design of work—test of a theory.

Organizational Behavior and Human Performance, 16: 250-279.

Heinze, T., Shapira, P., Rogers, J. D., Senker, J. M. 2009. Organizational and institutional

influences on creativity in scientific research. Research Policy, 38: 610-623.

Hollingsworth, R., Hollingsworth, E. J. 2000. Major discoveries and biomedical research

organizations: Perspectives on interdisciplinarity, nurturing leadership, and integrated

structure and cultures. In Weingart, P. & Stehr, N. (Eds.), Practising

interdisciplinarity: 215-244. Toronto, Canada: University of Toronto Press.

Kneller, R. 2007. Prospective and retrospective evaluation systems in context: Insights from

japan. In Whitley, R. & Glaser, J. (Eds.), Changing governance of the sciences: The

advent of research evaluation systems. Dordrecht, Netherlands.: Springer.

Knorr-Cetina, K. 1999. Epistemic cultures : How the sciences make knowledge. Cambridge,

MA: Harvard University Press.

Latour, B., Woolgar, S. 1979. Laboratory life: The construction of scientific facts. Princeton,

NJ: Princeton University Press.

Levin, S. G., Stephan, P. E. 1989. Age and research productivity of academic scientists.

Research in Higher Education, 30: 531-549.

Lindsey, D. 1980. Production and citation measures in the sociology of science - the problem of

multiple authorship. Social Studies of Science, 10: 145-162.

Long, J. S., McGinnis, R. 1981. Organizational context and scientific productivity. American

Sociological Review, 46: 422-442.

Merton, R. K. 1968. The bearing of empirical research on sociological theory, Social theory and

social structure: 156-171. Glencoe, IL: The Free Press.

Merton, R. K. 1973. Sociology of science. Chicago: University of Chicago Press.

Merton, R. K., Barber, E. 2004. The travels and adventures of serendipity. A study in

sociological semantics and the sociology of science. Princeton: Princeton University

Press.

Narin, F., Pinski, G., Gee, H. H. 1976. Structure of biomedical literature. Journal of the

29

American Society For Information Science, 27: 25-45.

Nelson, R. R. 1959. The simple economics of basic scientific-research. Journal of Political

Economy, 67: 297-306.

Nightingale, P. 1998. A cognitive model of innovation. Research Policy, 27: 689-709.

Oecd. 2013. Key findings of the oecd-knowinno project on the careers of doctorate holders, Vol.

FP7-Adhoc-2007-13. Paris: OECD.

Owen-Smith, J. 2001. Managing laboratory work through skepticism: Processes of evaluation

and control. American Sociological Review, 66: 427-452.

Pelz, D. C., Andrews, F. M. 1966. Scientists in organizations. New York: Wiley.

Roach, M., Sauermann, H. 2010. A taste for science? Phd scientists' academic orientation and

self-selection into research careers in industry. Research Policy, 39: 422-434.

Salonius, A. 2008. Working in the lab: Social organization of research and training in

biomedical research labs in canada and its relationship to research funding

MCGILL UNIVERSITY, Quebec, Canada.

Sauermann, H., Cohen, W. M. 2010. What makes them tick? Employee motives and firm

innovation. Management Science, 56: 2134-2153.

Sauermann, H., Stephan, P. E. 2012. Conflicting logics? A multidimensional view of industrial

and academic science. Organization Science, forthcoming.

Seymore, S. B. 2009. Serendipity. North Carolina Law Review, 88: 185-211.

Shibayama, S. 2011. Distribution of academic research funds: A case of japanese national

research grant. Scientometrics, 88: 43-60.

Shibayama, S., Walsh, J. P., Baba, Y. 2012. Academic entrepreneurship and exchange of

scientific resources: Material transfer in life sciences and materials science in japanese

universities. American Sociological Review, 77: 804-830.

Shimizu, H., Nirei, M., Maruyama, K. 2012. Management of science, serendipity, and research

performance: Evidence from scientists' survey in the us and japan, Schumpeter

Conference. Brisbane, Australia.

30

Simon, H. A. 1957. Administrative behavior: A study of decision-making processes in

administrative organrzation. New York: Macmilian.

Stapleton, D. H. 2004. Creating a tradition of biomedical research. Contributions to the history

of the rockefeller university. New York: Rockefeller University Press.

Stephan, P. E. 1996. The economics of science. Journal of Economic Literature, 34: 1199-1235.

Stokes, D. E. 1997. Pasteurs quadrant: Basic science and technological innovation

Washington, D.C.: Brookings Institution Press.

Teasley, S. D., Covi, L. A., Krishnan, M. S., Olson, J. S. 2002. Rapid software development

through team collocation. Ieee Transactions on Software Engineering, 28: 671-683.

Traweek, S. 1988. Beamtimes and lifetimes: The world of high energy physicists. Cambridge,

MA: Harvard University Press.

Tunzelmann, N. V., Ranga, M., Martin, B., Geuna, A. 2003. The effects of size on research

performance: A spru review. Brighton, UK: SPRU.

Van Angel, P. 1992. Serendipity: Expect also the unexpected. Creativity and Innovation

Management, 1: 20-32.

Weesie, J. 1999. Seemingly unrelated estimation and the cluster-adjusted sandwich estimator.

Stata Technical Bulletin, 52: 34-47.

Whitley, R. 1984. The intellectual and social organization of the sciences. New York: Oxford

University Press.

Woodman, R. W., Sawyer, J. E., Griffin, R. W. 1993. Toward a theory of organizational creativity.

Academy of Management Review, 18: 293-321.

Zuckermann, H. 1977. Scientific elite: Nobel laureates in the united states. New York: Free

Press.

Figure 1 Research Process and Task Allocation

Figure 2 Prediction of Task Allocation Effect a

a LH stands for lab heads. Based on regression results (Models 3 and 4 in Tables 3 and 4), we predicted the lab productivity in terms of per-staff citation count

(top row) and pub count (bottom row) for basic and applied laboratories. In each phase, we compare two extreme patterns of task allocation. In planning, since a

lab head usually plays the leading role, a lab head’s solo leading vs. co-leading with members is of the primary interest. Similarly, in execution, members’ solo

leading vs. co-leading with a lab head is of concern. In writing, since a lab head’s and members’ roles are negatively correlated, a lab head’s solo leading vs.

members’ solo leading are compared. For the prediction, the mean values are used for all variables except the focal task allocation variables.

Table 1 Descriptive Statistics and Correlation of Variables a

a N=305. Bold italic: p<0.05.

Mean Std.Dev. Min Max 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

1 ln(Per-staff pub count) 1.608 .640 .000 3.135

2 ln(Per-staff citation count) 3.273 1.166 .000 6.201 .831

3 ln(Pre-tenure pub count) 1.535 .807 .000 3.714 .466 .438

4 ln(Pre-tenure citation count) 2.808 1.683 .000 7.140 .295 .417 .803

5 Time for research 3.770 1.535 1.000 6.000 -.007 .085 .037 .114

6 Foreign experience 3.633 1.548 1.000 6.000 -.062 -.045 -.106 -.039 .084

7 Inbred .105 .307 .000 1.000 .104 .036 .157 .071 -.067 -.071

8 Medical doctor .174 .380 .000 1.000 -.087 .054 -.032 .115 .068 .120 -.016

9 Female .033 .178 .000 1.000 -.151 -.180 -.086 -.029 .148 .103 -.063 .061

10 Budget/#staff 3.960 2.882 .278 17.500 -.024 .136 -.005 .062 .242 .028 -.045 .083 -.059

11 Lab age 13.214 7.747 1.000 35.000 .146 -.004 .114 -.071 -.165 .016 .384 -.049 -.065 -.140

12 Lab size 9.246 5.204 3.000 45.000 -.012 .173 .036 .132 .075 .059 .027 .265 -0.06 .131 .023

13 Top7 univ .489 .501 .000 1.000 .050 .102 .135 .170 .096 -.039 .136 -.050 -0.03 .116 -.042 .199

14 Basic research .540 .499 .000 1.000 -.132 .033 .013 .116 .166 -.013 .016 -.054 0.02 -.023 -.097 .037 .222

15 Members' planning 1.179 .489 .000 2.000 .055 .132 .035 .084 .006 -.065 .029 -.051 -.087 .095 .065 .132 .150 .035

16 Lab head's execution 1.227 .567 .000 2.000 -.048 .025 -.078 -.067 .005 .115 .080 .130 -.009 .052 .094 .022 .028 .037 .074

17 Members' writing (& lab head's not writing) .985 .407 .000 1.667 .072 -.053 .035 -.055 -.236 -.153 .110 -.154 -.099 -.057 .160 .017 .026 -.163 .359 -.093

Table 2 Patterns of Task Allocation a

a YES: main role and NO: otherwise. We computed the mean of the extent of engagement (0: none, 1: minor, and 2:

main role) in related tasks and ranks, and assigned YES if it is 1.5 or greater. The number of laboratories categorized

into each pattern and the percentage of those laboratories (in parentheses) are shown for the whole sample as well as

for the breakdown by research areas and university ranks.

1 YES - - 99 (32%) 55 (34%) 41 (30%) 55 (37%) 44 (28%)

2 - YES - 148 (49%) 78 (48%) 67 (49%) 72 (48%) 76 (49%)

3 - - YES 163 (53%) 81 (50%) 79 (57%) 80 (54%) 83 (53%)

4 NO NO NO 62 (20%) 30 (19%) 32 (23%) 27 (18%) 35 (22%)

5 NO NO YES 45 (15%) 22 (14%) 23 (17%) 23 (15%) 22 (14%)

6 NO YES NO 61 (20%) 38 (23%) 21 (15%) 31 (21%) 30 (19%)

7 NO YES YES 38 (12%) 17 (10%) 21 (15%) 13 (9%) 25 (16%)

8 YES NO NO 11 (4%) 6 (4%) 5 (4%) 5 (3%) 6 (4%)

9 YES NO YES 39 (13%) 26 (16%) 11 (8%) 22 (15%) 17 (11%)

10 YES YES NO 8 (3%) 7 (4%) 1 (1%) 6 (4%) 2 (1%)

11 YES YES YES 41 (13%) 16 (10%) 24 (17%) 22 (15%) 19 (12%)

Top 7

(N = 149)

Not Top 7

(N = 156)

Research Areas University RankMembers'

planning

Lab head's

execution

Members'

writing

All Sample

(N = 305)Basic

(N = 162)

Applied

(N = 138)

Table 3 Prediction of Scientific Productivity: Dependent variable = ln(Per-staff

citation count) a

a Unstandardized coefficients (standard errors in parentheses). Two-tailed test. † p<0.10; * p<0.05; ** p<0.01;***

p<0.001. Ordinary least squared.

Control variable

ln(Pre-tenure citation count) .249 *** (.039) .228 *** (.039) .215 *** (.057) .224 *** (.055)

Time for research .025 (.045) .024 (.044) .032 (.063) .034 (.067)

Foreign experience -.026 (.043) -.060 (.043) -.109 † (.063) .005 (.062)

Inbred .078 (.230) .151 (.225) .571 † (.330) -.344 (.314)

Medical doctor .064 (.207) .024 (.203) .092 (.305) -.202 (.291)

Female -1.083 ** (.355) -1.026 ** (.345) -1.375 ** (.484) -.673 (.581)

Budget/#staff .011 (.024) .008 (.023) .058 (.037) -.031 (.030)

Lab age .005 (.010) .006 (.010) -.014 (.014) .028 † (.015)

Top7 univ .020 (.135) -.052 (.132) -.304 (.197) .176 (.185)

Basic Research .016 (.151) -.812 † (.485)

Task allocation

Members' planning .241 † (.145) .226 (.210) .476 * (.213) .280 (.203)

Lab head's execution .011 (.117) -.418 ** (.160) .481 * (.189) -.385 * (.155)

Members' writing (& lab head's not writing) -.262 (.181) -.023 (.247) -.527 † (.274) .044 (.244)

Interaction

Members' planning x Basic research .162 (.283)

Lab head's execution x Basic research .895 *** (.232)

Members' writing x Basic research -.478 (.344)

F test 3.273 *** 3.672 *** 2.599 *** 2.701 ***

Log likelihood -405.684 -395.549 -213.330 -161.844

N 292 292 156 136

Basic labs Applied labs

Model 1 Model 3 Model 4Model 2

All labs

Table 4 Prediction of Scientific Productivity: Dependent variable = ln(Per-staff

pub count) a

a Unstandardized coefficients (standard errors in parentheses). Two-tailed test. † p<0.10; * p<0.05; ** p<0.01;***

p<0.001. Ordinary least squared.

Control variable

ln(Pre-tenure pub count) .337 *** (.043) .320 *** (.042) .315 *** (.060) .302 *** (.061)

Time for research .019 (.024) .017 (.023) .019 (.031) .037 (.039)

Foreign experience -.002 (.023) -.020 (.023) -.058 † (.032) .026 (.035)

Inbred .040 (.122) .081 (.119) .276 † (.162) -.193 (.182)

Medical doctor -.073 (.109) -.086 (.107) -.052 (.151) -.167 (.167)

Female -.311 † (.187) -.277 (.182) -.240 (.239) -.236 (.335)

Budget/#staff -.018 (.012) -.019 (.012) -.005 (.018) -.030 † (.017)

Lab age .006 (.005) .007 (.005) -.003 (.007) .020 * (.008)

Top7 univ -.007 (.071) -.047 (.070) -.155 (.097) .057 (.107)

Basic Research -.091 (.080) -.547 * (.255)

Task allocation

Members' planning .055 (.076) -.024 (.111) .219 * (.104) -.037 (.118)

Lab head's execution -.046 (.062) -.247 ** (.084) .173 † (.092) -.197 * (.088)

Members' writing (& lab head's not writing) -.035 (.096) .130 (.130) -.225 † (.135) .177 (.141)

Interaction

Members' planning x Basic research .206 (.150)

Lab head's execution x Basic research .429 *** (.122)

Members' writing x Basic research -.327 † (.182)

F test 4.166 *** 4.542 *** 2.779 *** 3.288 ***

Log likelihood -219.014 -208.888 -103.119 -87.426

N 292 292 156 136

Basic labs Applied labs

Model 1 Model 3 Model 4Model 2

All labs

Appendix 1 Factor Analysis of Task Allocation a

a Relatively large factor loadings, conceptually pertinent to each factor, are indicated in bold italic.

Factor1 Factor2 Factor3 Factor4 Factor5 Factor6

Members'

planning

Junior

researcher’s

full responsibility

Lab head’s

planning

PhD’s

execution

Lab head’s

execution

Members'

writing

Subject -.077 .133 .750 .002 -.032 -.088

Hypothesis -.020 .088 .836 .170 .087 .057

Planning .064 -.062 .579 -.007 .498 -.052

Experiment -.006 .046 -.019 -.055 .857 -.004

Analysis .070 .166 .259 .048 .677 -.125

Writing .167 .232 .348 .120 .282 -.629

Subject .666 .423 .104 -.121 .014 .087

Hypothesis .748 .446 .115 -.071 .015 -.051

Planning .544 .594 .107 .060 -.152 -.034

Experiment -.036 .811 .047 .115 .190 -.122

Analysis .154 .822 .035 .206 .029 -.118

Writing .174 .698 .138 .030 .012 .497

Subject .675 -.047 -.184 .163 .163 .268

Hypothesis .812 -.010 -.090 .323 .043 .104

Planning .698 -.028 -.045 .404 -.048 .117

Experiment .052 .144 .072 .884 .001 .046

Analysis .303 .125 .114 .787 -.035 .175

Writing .261 .003 .065 .273 .013 .825

PhD

Junior

researcher

Lab head